Nonvolatile resistive switching is one of the key phenomena for emerging applications in optoelectronics and neuromorphic computing. In most of the cases, an electric field is applied to a two terminal dielectric material device and leads to the formation of a low resistance filament due to ion migration. However, the stochastic nature of the ion migration can be an impediment for the device robustness and controllability, with uncontrolled variations of high and low resistance states or threshold voltages. Here, we report an optically induced resistive switching based on a CdS/V3O5 heterostructure which can overcome this issue. V3O5 is known to have a second order insulator to metal transition around Tc ≈ 415 K, with an electrically induced threshold switching at room temperature. Upon illumination, the direct transfer of the photoinduced carriers from the CdS into V3O5 produces a nonvolatile resistive switching at room temperature. The initial high resistance can be recovered by reaching the high temperature metallic phase, i.e., temperatures above Tc. Interestingly, this resistive switching becomes volatile around the Tc. By locally manipulating the volatile and nonvolatile resistive switching using electric field and light, this system is a promising platform for hardware based neuromorphic computing implementations.

Resistive switching is a very active research field and is developing rapidly due to its fundamental interest and potential applications, such as nonvolatile resistive random-access memories (RRAM),1 memristors,2 or neuromorphic computing.3–6 This phenomenon mostly refers to an electric field induced reversible, nonvolatile resistive switching of a dielectric material (like a transition-metal oxide) sandwiched between two electrodes. Under a voltage pulse (SET mechanism), the two-terminal device switches from a high resistance state (HRS) to a low resistance state (LRS) and remains in the LRS. The physical mechanism behind the resistive switching is generally ion migration forming oxygen-vacancy or metal-atom filaments bridging the electrodes.7 Successive electric field pulses can dissolve the filaments (RESET mechanism), bringing the device back to a HRS. On the other hand, strongly correlated materials, such as temperature-driven insulator to metal transition (IMT) compounds,8 can also exhibit an electrically induced resistive switching below the transition temperature (Tc). The most studied compounds are VO2 and V2O3. In these cases, the resistive switching is volatile (also called threshold switching) since the resistance of the material returns to its initial value once the electrical bias is removed. In contrast to the previous nonvolatile effect that involved ionic migration, it has been demonstrated on several IMT materials that this volatile resistive switching is a result of the formation of a voltage induced metallic filament.9,10 In addition, it was shown that the nonvolatile resistive switching could be induced in VO2 devices far from Tc, leading to the creation of a new vanadium oxide phase filament.11 So far, the electrically triggered resistive switching has been extensively studied in these strongly correlated materials. On the other hand, different optoelectronic resistive switching systems have been reported, where light induces threshold switching,12,13 enhances or modulates resistive switching and threshold switching,14–18 or the resistive switching is volatile.19 The optical control of the switching response has attracted much attention over the past few years.

Recently, we have shown that combining an ultra-thin film (10 nm) of V2O3 or VO2 with a photoresistor cadmium sulfide (CdS) layer allows for a modulation of the transition temperature and the resistance in the low temperature insulating phase upon visible light.20 CdS is a semiconductor that generates a high concentration of charge carriers (holes and electrons) when exposed to light.21 In the previous work,20 it was reported that illumination of a CdS/V2O3 heterostructure, under illumination of the transition, completely suppresses the transition, leading to a giant resistance decrease in at least six orders of magnitude. In the CdS/VO2 case, the resistance change was smaller but still remarkable, and the transition was not completely suppressed but shifted to lower temperatures by ∼15 K. The resistance modulation was volatile in the CdS/V2O3, i.e., it recovered its initial resistance when the illumination is stopped. However, the modulation is essentially nonvolatile in the case of the CdS/VO2 heterostructure, i.e., the change in resistance R(T) under illumination remained the same after the illumination was removed. However, after 8 h at 300 K, the heterostructure recovered its initial R(T), thus reversing the modulation effect. Therefore, the switching mechanism may be different between different CdS/oxides systems.

In this work, we focus on another vanadium oxide, V3O5, which, in contrast to VO2 and V2O3, exhibits a second order IMT around Tc ≈ 415 K. Here, we sputtered a CdS layer on top of a V3O5 thin film, and we studied the effect of the light on the resistance of this heterostructure. Our results show a nonvolatile resistive switching at room temperature upon illumination, i.e., 120 K below Tc. The resistive switching ratio is about 80%. As the temperature increases, the resistive switching ratio decreases (down to 32%), and the switching becomes volatile. Density-functional theory (DFT) calculations show that localization of trapped electron in naturally present oxygen vacancies is responsible for the nonvolatile resistive switching. Contrary to VO2 that does not relax even at high temperature, in the V3O5, the resistance returns to its initial state once the temperature is above Tc. As shown previously,22 the V3O5 high temperature phase can be reached by electrically triggering volatile resistive switching. Our results provide, therefore, a great platform for neuromorphic computing applications, especially using electrically induced switching. In addition, because in our structure, no ion migration is involved, these heterostructures overcome robustness and controllability issues arising from ion migration. Thus, CdS/V3O5 devices could be used to implement neuron-like (electrical bias) and synapse-like (light + electrical bias) functionalities in the same material system.

Figure 1(a) shows the changes in the resistance vs temperature [R(T)] when illuminating the CdS/V3O5 hybrid [Fig. 1(a)] compared to R(T) in the dark. Without illumination [black (heating 1—cooling 2)] curves in Fig. 1(a), the CdS/V3O5 hybrid exhibits a second order IMT around 415 K, as expected.23,24 This R(T) curve is similar to previous studies22,25 with a strong nonlinear temperature dependence within the insulating state (300–400 K) and a change of resistance during the phase transition (factor of 3 between 400 and 420 K). The illumination of the CdS/V3O5 hybrid with a white LED induces a resistive switching at room temperature of a factor of 5. A clear resistive switching can be observed when comparing the initial R(T) measured in the dark, with the R(T) measured upon a 481 mW/cm2 illumination at 300 K [black (heating 1—cooling 2) and orange (heating 3—cooling 4) curves, respectively]. We note that the resistance changes decrease with increasing the temperature. Illumination of a bare V3O5 thin film with the same power density shows no resistance variations, implying that this phenomenon is not caused by thermal heating. Interestingly, in VO2,18 the resistance change remains during both heating and cooling curves in the illuminated state even after switching off the light. On the other hand, the resistance of V3O5 increases slightly in the dark but does recover its original value around the phase transition temperature (dashed blue line—heating curve 5). The initial state is recovered in the cooling loop (dashed blue line—cooling curve 6). Therefore, in the CdS/V3O5 hybrid, illumination provides a SET mechanism, and the voltage induced transition into in the high temperature phase would provide the RESET mechanism.

FIG. 1.

(a) Resistance of CdS/V3O5 hybrid as a function of temperature. Black and orange curves correspond to the dark and 481 mW/cm2 illumination, respectively. Dashed curves (5 and 6) obtained after illumination. Arrows indicate sweep directions. (b) Time dependence of the same hybrid at 300 K. The resistance normalized to before illumination. At t =0, 481 mW/cm2 illumination started that abruptly changes the resistance by a factor of 5 until turned off at t 3 min. The inset shows resistance relaxation for 6 h.

FIG. 1.

(a) Resistance of CdS/V3O5 hybrid as a function of temperature. Black and orange curves correspond to the dark and 481 mW/cm2 illumination, respectively. Dashed curves (5 and 6) obtained after illumination. Arrows indicate sweep directions. (b) Time dependence of the same hybrid at 300 K. The resistance normalized to before illumination. At t =0, 481 mW/cm2 illumination started that abruptly changes the resistance by a factor of 5 until turned off at t 3 min. The inset shows resistance relaxation for 6 h.

Close modal

The CdS/V3O5 hybrid shows strong resistive switching when illuminated. Figure 1(b) shows the time dependence of the resistance normalized to the insulating pristine state (before illumination) at 300 K. At t =0, the light is switched on, illuminating the hybrid with a 481 mW/cm2 power density. The resistance of the V3O5 (R/Rdark) abruptly changes by a factor of 5. After removing the light, the resistance slightly increases but remains in a low resistance value. Our measurement shows that after 6 h, the V3O5 only recovered 32% of its initial state, i.e., the relaxation of the resistive switching ratio [ΔR = (Rdark − Rill)/Rdark] is around 17%. A direct measurement of the full recovery time is too long to be feasible. From a fit of the short time recovery of the resistance (see the supplementary material), we estimate that a 50% recovery of the initial resistance would take more than 10 years. So, in order to accelerate the time to recover Rdark, the heterostructures are heated above Tc.

Figure 2(a) shows the time dependence of the normalized resistance change at different temperatures. The amplitude and the nature (volatile vs nonvolatile) of the resistive switching are strongly dependent on the temperature. For temperatures below Tc (< 400 K), the resistive switching ratio (ΔR) decreases with increasing the temperature [Fig. 2(c)]. At 400 K (near Tc), however, the resistive switching ratio becomes non-monotonic and is more pronounced than at 360 and 380 K, suggesting a change of regime near the onset of the phase transition. The relaxation of the resistance after illumination [Fig. 2(b)] fits well to a stretched exponential as shown in the supplementary material. Far below the phase transition, the resistive switching is “non-volatile,” i.e., the resistance relaxes only slightly. 1 h after the LED was turned OFF, the resistance ratio only relaxed by 10% [Fig. 2(c)]. We found that the higher the temperature, the smaller the non-volatility effect of the resistive switching. The V3O5 has, indeed, relaxed by 39% of its initial resistance at 340 K, 56% at 360 K, 72% at 380 K, and 89% at 400 K. Hence, the resistive switching that is nonvolatile at room temperature becomes volatile at higher temperature, just beneath the onset of the phase transition.

FIG. 2.

(a) Time dependence of the resistance normalized to the insulating state at different temperatures. At t =0, the sample is illuminated, and the system switches into a low resistance state. (b) Time dependence of the resistance normalized to the insulating state. At t =0, the LED is switched OFF. (c) Temperature dependence of resistive switching ratio (ΔR = (Rdark − Rill)/Rdark) (orange triangle) and relaxation of ΔR (ΔR relaxation = (ΔR − ΔRafter illumination)/ΔR) (black dots) after 1 h.

FIG. 2.

(a) Time dependence of the resistance normalized to the insulating state at different temperatures. At t =0, the sample is illuminated, and the system switches into a low resistance state. (b) Time dependence of the resistance normalized to the insulating state. At t =0, the LED is switched OFF. (c) Temperature dependence of resistive switching ratio (ΔR = (Rdark − Rill)/Rdark) (orange triangle) and relaxation of ΔR (ΔR relaxation = (ΔR − ΔRafter illumination)/ΔR) (black dots) after 1 h.

Close modal

To elucidate the photoinduced effect, we performed DFT calculations to understand the defects present in V3O5 and the effects of electron doping. We find that oxygen vacancies are the most stable defects in the case of LT (low temperature) phase (Fig. S1), which has also been reported experimentally by Rua et al.26 The methods used for the defect calculations are presented in the supplementary material, and the various defect energies as a function of oxygen chemical potential are shown in Fig. S1.

DFT calculations were performed on supercells with negatively charged oxygen vacancies with varying electron density. We note that for computational tractability, the calculations utilized a single oxygen vacancy within a 2 × 2 × 2 supercell (96 formula units), which significantly overestimates the density of oxygen vacancies. Nevertheless, we expect that the qualitative trends are still valid. With a charged oxygen vacancy, the bandgap decreases from 0.63 to 0.40 eV [Figs. 3(e) and 3(f); a wider DOS is shown in the supplementary material]. The Density of State (DOS) for the system with negatively charged oxygen vacancies reveals the appearance of a small peak near EF, in the forbidden zone [Fig. 3(f)]. This suggests a deep state with electron localization. The localization can be further confirmed from the atomic DOS for vanadium sites present in the near vicinity of the oxygen vacant site, where the same small peak is observed near the EF (Fig. S2). The same phenomenon can also be observed from the wavefunction plots in Fig. S3(c). This peak vanishes for vanadium sites further away from the oxygen vacancy. We note that this localization is observed only with negatively charged oxygen vacancies, indicating no possibility of hole doping in the system. Figure S5 shows that the hole doped system is metallic as the EF is inside the valence band. Moreover, the presence of small peak just below the EF confirms that the electrons (generated by CdS under illumination) are injected in V3O5 and are trapped into this deep state, slowing the “low” temperature relaxation rate. For the LT phase (T <340 K), the photoexcited electrons trapped into the oxygen vacancy are unable to overcome the barrier, and therefore, no relaxation is observed. At higher temperatures, relaxation is induced by thermal excitation of the trapped electron over the barrier and via the thermally accessible conduction band.

FIG. 3.

(a)–(d) Schematics proposed mechanism for the nonvolatile resistive switching in the CdS/V3O5 hybrid. (a) Band structure of V3O5 with in-gap state due to the presence of oxygen vacancies. (b) Under illumination, electrons from CdS are doping the V3O5, and filling the oxygen vacancies in gap state. (c) Once the illumination is removed, the system is no longer doped but remains in the low resistance state since electrons are trapped in oxygen vacancies. (d) At temperatures above Tc, the gap thermally closes, and electrons trapped into the oxygen vacancies are released. The system then goes back to the case a), where the Fermi level is low and far from the conduction band. (e) and (f) Total (green), O-2p (red), and V-3d (blue) orbital density of states (DOS) for a 2 × 2 × 2 supercell of (e) LT stoichiometric V3O5 and (f) nonstoichiometric V3O5 containing a negatively charged oxygen vacancy. A decrease in the bandgap can be seen from 0.63 eV in (e) to 0.40 eV in (f).

FIG. 3.

(a)–(d) Schematics proposed mechanism for the nonvolatile resistive switching in the CdS/V3O5 hybrid. (a) Band structure of V3O5 with in-gap state due to the presence of oxygen vacancies. (b) Under illumination, electrons from CdS are doping the V3O5, and filling the oxygen vacancies in gap state. (c) Once the illumination is removed, the system is no longer doped but remains in the low resistance state since electrons are trapped in oxygen vacancies. (d) At temperatures above Tc, the gap thermally closes, and electrons trapped into the oxygen vacancies are released. The system then goes back to the case a), where the Fermi level is low and far from the conduction band. (e) and (f) Total (green), O-2p (red), and V-3d (blue) orbital density of states (DOS) for a 2 × 2 × 2 supercell of (e) LT stoichiometric V3O5 and (f) nonstoichiometric V3O5 containing a negatively charged oxygen vacancy. A decrease in the bandgap can be seen from 0.63 eV in (e) to 0.40 eV in (f).

Close modal

The above results imply that the nonvolatile resistive switching is due to charge-transfer of the photoinduced carrier.27,28 The photon absorbed in the CdS excites an electron into the conduction band, and this has a finite probability to be transferred into the V3O5. The electron doping in oxygen-deficient V3O5 produces a decrease in the resistance by decreasing the gap as well as filling the in-gap state produced by the oxygen vacancies [Fig. 3(b)]. Illumination decreases the Fermi level. Although some carriers in the higher in-gap states may thermally relax, most of the carrier remain trapped in the in-gap state, unable to overcome the energy barrier. As a consequence, the system remains in the low resistance state [Fig. 3(c)]. At higher temperature, when reaching the phase transition, the gap is thermally closing, and the thermal excitation is more important. This results in the relaxation of trapped carrier, either by thermal excitation or via the thermally accessible conductions bands.

In conclusion, we have demonstrated illumination induced room temperature nonvolatile resistive switching in a CdS/V3O5 heterostructure. This light-induced resistive switching can be reset at higher temperature at T >415 K. Although the resistive switching is nonvolatile at temperatures far below Tc, it becomes volatile when close to the transition (T >380 K). The relaxation of the resistance after illumination has been fitted with stretch exponentials, and the extracted thermally activated relaxation times correspond to an energy barrier close to 847 meV. Contrary to the common electrically induced filamentary resistive switching observed in many simple oxides, the resistive switching presented here is due to photodoping, with carriers created in the CdS during the illumination and injected into the V3O5. DFT calculations support the idea that photoexcited electrons can be trapped into oxygen vacancies and are unable to overcome the energy barrier at room temperature. Moreover, this non-volatility could be incorporated with volatile resistive switching (threshold switching)22 in simple V3O5 two-terminal device based systems. The neuron-like functionalities could be emulated by electrical excitation, and the coupling of light (SET) and electrical (RESET) excitations would lead to synapse-like functions. This opens the use of a new optoelectronic type memristor for emergent neuromorphic computing.

The V3O5 film was synthetized by transforming a V2O3 (60 nm) film grown on ⟨012⟩ Al2O3 through a gas evolution technique.25 This system allows for precise control of the oxygen partial pressure and temperature starting from high vacuum conditions. A thorough description of this technique and a detailed investigation of the electrical, structural, and compositional properties of the V3O5 thin film can be found in our previous paper.25 The sample was annealed at 837 °C under an oxygen partial pressure ∼9 × 10−8  Torr for 6 h, until the oxidation from V2O3 into V3O5 was complete. θ–2θ XRD and reciprocal space maps together with a sharp metal-insulator transition indicate the presence of single phase V3O5 thin films after annealing. On top of these films, we deposited an 80 nm thick CdS film using rf magnetron sputtering, in a 2-mTorr pure argon atmosphere at 180 °C. Two 500 μm wide electrodes were pattern in a coplanar geometry and were separated with a 6 mm distance on top of the CdS/V3O5 heterostructures. XRD measurements were performed in a Rigaku SmartLab system at room temperature. Single-phase growth is confirmed by XRD, epitaxially along the ⟨012⟩ direction for V2O3. After the transformation process, V3O5 and CdS become polycrystalline. Transport measurements were carried out on TTPX Lakeshore cryogenic probe station, using a Keithley 6221 current source and a Keithley 2182A nanovoltmeter. We used a Thorlabs LED with 400–700 nm wavelength range during the transport measurements.

All spin-polarized DFT29,30 calculations were performed using the Vienna Ab initio Simulation Package (VASP)29,31,32 within the projector augmented wave (PAW)33 approach. The exchange-correlation effects were modeled using the Perdew–Burke–Ernzerhof (PBE)34 generalized gradient approximations (GGA)34 functional with a plane wave cutoff of 520 eV. The pseudopotentials used were similar to those used in the Materials Project,35 where the core electrons were described with the frozen core approximation by the PAW potential, while the valence electrons of vanadium and oxygen are given by V:3p6 3d4 4s1 and O:2s2 2p4, respectively. The convergence criteria were set to 1 × 10−4 eV and −0.02 eV/Å for energies and atomic forces, respectively. The tetrahedron method with Blöchl corrections36 was used for the insulating LT phase with Γ-centered k-point meshes of 1 × 2 × 2 and 3 × 4 × 4 for the relaxation of the supercells and for DOS calculations, respectively. All analysis and input generation were performed with the Python Materials Genomics (pymatgen) library.37 

The initial V3O5 crystal structure with a space group P2/c was taken from Materials Project. The DFT + U method38 was used for all calculations, with a Hubbard parameter U of 3.5 eV for V.39 This U value resulted in a bandgap of 0.63 eV for LT V3O5, reproducing the experimental bandgap.40 All calculations were performed in a ferromagnetic configuration.

All defect calculations were performed using 2 × 2 × 2 (256 atoms) supercells of the conventional unit cell for the LT phase, which is large enough to minimize interactions between the periodic boundary images. We investigated neutral point defects focusing on vacancies and interstitials for both oxygen and vanadium. The vacancies were formed by removing one atom from the supercell and allowing the remaining atoms to relax. Interstitials were created by adding an atom at one of the ten symmetrically distinct sites found using Interstitially Finding Tool (InFiT).41 More details of this method can be found in the work by Broberg et al.42 Charged defects were studied by electron doping via the addition of integer numbers of electrons. It should be noted that VASP automatically adds a background compensating charge to achieve an overall charge neutral unit cell.

See the supplementary material related to the thermodynamic analysis and the structural characterization of the analyzed heterostructure.

The experimental investigation of coupled photoconducting and Mott materials was supported by the U.S. Air Force Office of Scientific Research under Award No. FA9550-20-1-0242. The theoretical aspects of this work were supported as part of the Quantum Materials for Energy Efficient Neuromorphic Computing (Q-MEEN-C), an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Basic Energy Sciences, under Award No. DE-SC0019273. The authors also acknowledge computational resources provided by the Triton Shared Computing Cluster (TSCC) at the University of California San Diego. This research used resources of the National Energy Research Scientific Computing Center (NERSC), a U.S. Department of Energy Office of Science User Facility located at Lawrence Berkeley National Laboratory, operated under Contract No. DE-AC02-05CH11231.

The authors have no conflicts to disclose.

Coline Adda: Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review and editing (equal). Henry Navarro: Data curation (lead); Formal analysis (lead); Investigation (lead); Methodology (lead); Writing – original draft (lead); Writing – review and editing (equal). Jasleen Kaur: Resources (equal); Software (equal). Min-Han Lee: Investigation (equal); Resources (equal). Chi Chen: Resources (equal); Software (equal). Marcelo J. Rozenberg: Conceptualization (equal); Investigation (equal); Validation (equal); Writing – review and editing (equal). Shyue Ping Ong: Investigation (equal); Methodology (equal); Software (lead); Writing – original draft (equal); Writing – review and editing (equal). Ivan K. Schuller: Funding acquisition (equal); Investigation (equal); Resources (equal); Supervision (lead); Writing – original draft (equal); Writing – review and editing (lead).

The data that support the findings of this study are available from the corresponding authors upon reasonable request.

1.
S.
Larentis
,
F.
Nardi
,
S.
Balatti
,
D. C.
Gilmer
, and
D.
Ielmini
,
IEEE Trans. Electron Devices
59
(
9
),
2468
2475
(
2012
).
2.
S.
Kim
,
S.
Choi
, and
W.
Lu
,
ACS Nano
8
(
3
),
2369
2376
(
2014
).
3.
Z.
Wang
,
H.
Wu
,
G. W.
Burr
,
C. S.
Hwang
,
K. L.
Wang
,
Q.
Xia
, and
J. J.
Yang
,
Nat. Rev. Mater.
5
(
3
),
173
195
(
2020
).
4.
J. S.
Lee
,
S.
Lee
, and
T. W.
Noh
,
Appl. Phys. Rev.
2
(
3
),
031303
(
2015
).
5.
Y.
Li
,
Z.
Wang
,
R.
Midya
,
Q.
Xia
, and
J. J.
Yang
,
J. Phys. D
51
(
50
),
503002
(
2018
).
6.
R.
Rocco
,
J.
del Valle
,
H.
Navarro
,
P.
Salev
,
I. K.
Schuller
, and
M.
Rozenberg
,
Phys. Rev. Appl.
17
(
2
),
024028
(
2022
).
7.
D.
Lelmini
and
R.
Waser
,
Resistive Switching: From Fundamentals of Nanoionic Redox Processes to Memristive Device Applications
(
Wiley-VCH Verlag GmbH & Co. KGaA
,
2016
).
8.
N. F.
Mott
,
Rev. Mod. Phys.
40
(
4
),
677
683
(
1968
).
9.
A.
Zimmers
,
L.
Aigouy
,
M.
Mortier
,
A.
Sharoni
,
S.
Wang
,
K. G.
West
,
J. G.
Ramirez
, and
I. K.
Schuller
,
Phys. Rev. Lett.
110
(
5
),
056601
(
2013
).
10.
S.
Kumar
,
M. D.
Pickett
,
J. P.
Strachan
,
G.
Gibson
,
Y.
Nishi
, and
R. S.
Williams
,
Adv. Mater.
25
(
42
),
6128
6132
(
2013
).
11.
S.
Cheng
,
M.-H.
Lee
,
X.
Li
,
L.
Fratino
,
F.
Tesler
,
M.-G.
Han
,
J.
del Valle
,
R. C.
Dynes
,
M. J.
Rozenberg
,
I. K.
Schuller
, and
Y.
Zhu
,
Proc. Natl. Acad. Sci. U. S. A.
118
(
9
),
e2013676118
(
2021
).
12.
S.
Stathopoulos
,
I.
Tzouvadaki
, and
T.
Prodromakis
,
Sci. Rep.
10
(
1
),
21130
(
2020
).
13.
A.
Kathalingam
,
H.-S.
Kim
,
S.-D.
Kim
, and
H.-C.
Park
,
Opt. Mater.
48
,
190
197
(
2015
).
14.
B.
Sun
,
Y.
Liu
,
W.
Zhao
, and
P.
Chen
,
RSC Adv.
5
(
18
),
13513
13518
(
2015
).
15.
M.
Ungureanu
,
R.
Zazpe
,
F.
Golmar
,
P.
Stoliar
,
R.
Llopis
,
F.
Casanova
, and
L. E.
Hueso
,
Adv. Mater.
24
(
18
),
2496
2500
(
2012
).
16.
X.
Guan
,
W.
Hu
,
M. A.
Haque
,
N.
Wei
,
Z.
Liu
,
A.
Chen
, and
T.
Wu
,
Adv. Funct. Mater.
28
(
3
),
1704665
(
2018
).
17.
S.
Sharma
,
A.
Kumar
, and
D.
Kaur
,
Appl. Phys. Lett.
115
(
5
),
052108
(
2019
).
18.
A.
Mehonic
,
T.
Gerard
, and
A. J.
Kenyon
,
Appl. Phys. Lett.
111
(
23
),
233502
(
2017
).
19.
H.-K.
He
,
R.
Yang
,
W.
Zhou
,
H.-M.
Huang
,
J.
Xiong
,
L.
Gan
,
T.-Y.
Zhai
, and
X.
Guo
,
Small
14
(
15
),
1800079
(
2018
).
20.
H.
Navarro
,
J.
del Valle
,
Y.
Kalcheim
,
N. M.
Vargas
,
C.
Adda
,
M.-H.
Lee
,
P.
Lapa
,
A.
Rivera-Calzada
,
I. A.
Zaluzhnyy
,
E.
Qiu
,
O.
Shpyrko
,
M.
Rozenberg
,
A.
Frano
, and
I. K.
Schuller
,
Appl. Phys. Lett.
118
(
14
),
141901
(
2021
).
21.
S.
Kose
,
F.
Atay
,
V.
Bilgin
,
I.
Akyuz
, and
E.
Ketenci
,
Appl. Surf. Sci.
256
(
13
),
4299
4303
(
2010
).
22.
C.
Adda
,
M.-H.
Lee
,
Y.
Kalcheim
,
P.
Salev
,
R.
Rocco
,
N. M.
Vargas
,
N.
Ghazikhanian
,
C.-P.
Li
,
G.
Albright
,
M.
Rozenberg
, and
I. K.
Schuller
,
Phys. Rev. X
12
(
1
),
011025
(
2022
).
23.
N.
Kumar
,
A.
Rúa
,
J.
Lu
,
F.
Fernández
, and
S.
Lysenko
,
Phys. Rev. Lett.
119
(
5
),
057602
(
2017
).
24.
V. N.
Andreev
and
V. A.
Klimov
,
Phys. Solid State
53
(
12
),
2424
2430
(
2011
).
25.
M.-H.
Lee
,
Y.
Kalcheim
,
J.
del Valle
, and
I. K.
Schuller
,
ACS Appl. Mater. Interfaces
13
(
1
),
887
896
(
2021
).
26.
A.
Rúa
,
R. D.
Díaz
,
N.
Kumar
,
S.
Lysenko
, and
F. E.
Fernández
,
J. Appl. Phys.
121
(
23
),
235302
(
2017
).
27.
V. I.
Kudinov
,
I. L.
Chaplygin
,
A. I.
Kirilyuk
,
N. M.
Kreines
,
R.
Laiho
,
E.
Lähderanta
, and
C.
Ayache
,
Phys. Rev. B
47
(
14
),
9017
9028
(
1993
).
28.
G.
Nieva
,
E.
Osquiguil
,
J.
Guimpel
,
M.
Maenhoudt
,
B.
Wuyts
,
Y.
Bruynseraede
,
M. B.
Maple
, and
I. K.
Schuller
,
Appl. Phys. Lett.
60
(
17
),
2159
2161
(
1992
).
29.
G.
Kresse
and
J.
Furthmüller
,
Phys. Rev. B
54
(
16
),
11169
11186
(
1996
).
30.
P.
Hohenberg
and
W.
Kohn
,
Phys. Rev.
136
(
3B
),
B864
B871
(
1964
).
31.
G.
Kresse
and
J.
Hafner
,
Phys. Rev. B
47
(
1
),
558
561
(
1993
).
32.
G.
Kresse
and
J.
Furthmüller
,
Comput. Mater. Sci.
6
(
1
),
15
50
(
1996
).
33.
P. E.
Blöchl
,
Phys. Rev. B
50
(
24
),
17953
17979
(
1994
).
34.
J. P.
Perdew
,
K.
Burke
, and
M.
Ernzerhof
,
Phys. Rev. Lett.
77
(
18
),
3865
3868
(
1996
).
35.
A.
Jain
,
S. P.
Ong
,
G.
Hautier
,
W.
Chen
,
W. D.
Richards
,
S.
Dacek
,
S.
Cholia
,
D.
Gunter
,
D.
Skinner
,
G.
Ceder
, and
K. A.
Persson
,
APL Mater.
1
(
1
),
011002
(
2013
).
36.
P. E.
Blöchl
,
O.
Jepsen
, and
O. K.
Andersen
,
Phys. Rev. B
49
(
23
),
16223
16233
(
1994
).
37.
S. P.
Ong
,
W. D.
Richards
,
A.
Jain
,
G.
Hautier
,
M.
Kocher
,
S.
Cholia
,
D.
Gunter
,
V. L.
Chevrier
,
K. A.
Persson
, and
G.
Ceder
,
Comput. Mater. Sci.
68
,
314
319
(
2013
).
38.
V. I.
Anisimov
,
J.
Zaanen
, and
O. K.
Andersen
,
Phys. Rev. B
44
(
3
),
943
954
(
1991
).
39.
B.
Himmetoglu
,
A.
Floris
,
S.
de Gironcoli
, and
M.
Cococcioni
,
Int. J. Quantum Chem.
114
(
1
),
14
49
(
2014
).
40.
N. J.
Szymanski
,
Z. T. Y.
Liu
,
T.
Alderson
,
N. J.
Podraza
,
P.
Sarin
, and
S. V.
Khare
,
Comput. Mater. Sci.
146
,
310
318
(
2018
).
41.
N. E. R.
Zimmermann
,
M. K.
Horton
,
A.
Jain
, and
M.
Haranczyk
,
Front. Mater.
4
,
34
(
2017
).
42.
D.
Broberg
,
B.
Medasani
,
N. E. R.
Zimmermann
,
G.
Yu
,
A.
Canning
,
M.
Haranczyk
,
M.
Asta
, and
G.
Hautier
,
Comput. Phys. Commun.
226
,
165
179
(
2018
).

Supplementary Material